Currently, in the era of mass data, large amounts of data are stored in storage in the form of readable file, such as ASCII code, XML, UNICODE, etc. When processing these data, the traditional method is: the CPU (Central Processing Unit) of server firstly reads the file from storage and writes into memory, then converts the data in memory and converts the ASCII code into a binary number that could be operated. However, the traditional method of data conversion consumes a large amount of resources of CPU and DRAM (Dynamic Random Access Memory), and it takes a lot of processing time. At the same time, the CPU has to frequently perform process switching and interrupt handling in the process of performing data conversion, consuming a lot of time and resources. According to statistics, about 60% of the time of each data processing business will spend on the data conversion, and only 40% of the time is actually used in operation.
Recently, it was proposed to use an embedded processor of SSD (Solid State Drives) controller to perform data conversion by the way of software program. However, running program in the MCU (Microcontroller Unit) of the SSD controller not only consumes a large amount of limited resources of the controller, but also costs much longer time to perform data conversion than running program in the CPU of server, because the processing ability of the MCU is far from being so good as that of the CPU of server.
How to improve the efficiency of data conversion without occupying resources of the CPU and DRAM of the server or resources of the MCU of the SSD controller is a technical problem that we urgently need to solve.